Unusual ferromagnetism enhancement in ferromagnetically optimal manganite La0.7−yCa0.3+yMn1−yRuyO3 (0≤y<0.3): the role of Mn-Ru t2g super-exchange

The eg-orbital double-exchange mechanism as the core of physics of colossal magnetoresistance (CMR) manganites is well known, which usually covers up the role of super-exchange at the t2g-orbitals. The role of the double-exchange mechanism is maximized in La0.7Ca0.3MnO3, leading to the concurrent metal-insulator transition and ferromagnetic transition as well as CMR effect. In this work, by a set of synchronous Ru-substitution and Ca-substitution experiments on La0.7–yCa0.3+yMn1–yRuyO3, we demonstrate that the optimal ferromagnetism in La0.7Ca0.3MnO3 can be further enhanced. It is also found that the metal-insulator transition and magnetic transition can be separately modulated. By well-designed experimental schemes with which the Mn3+-Mn4+ double-exchange is damaged as weakly as possible, it is revealed that this ferromagnetism enhancement is attributed to the Mn-Ru t2g ferromagnetic super-exchange. The present work allows a platform on which the electro-transport and magnetism of rare-earth manganites can be controlled by means of the t2g-orbital physics of strongly correlated transition metal oxides.

but insignificant 7 . To some extent, the Jahn-Teller effect in LCMO is important but not so sensitive to intrinsic substitutions and external stimuli too 4 , and therefore will not be considered here for convenience of discussion. Upon the Ca 21 partial substitution of La 31 , the asgenerated Mn 41 ions coexist with Mn 31 ions, giving the Mn 41 /Mn 31 ratio g5x/1-x. The FM transition point T C reaches the maximal T C ,265K at x,0.3, at which the Mn 41 charge density d 41 is ,0.3 and the Mn 31 charge density d 31 is ,0.7. Further increasing of x leads to spatially ordered Mn 41 :Mn 31 charge-ordering (CO) and even orbital-ordering (OO) sequence [17][18][19][20][21][22][23] , which on the other hand favors the antiferromagnetic (AFM) insulating transitions at x,0.5 and above. Therefore, the optimal ferromagnetism and best metallic conductivity in LCMO appears at x50. 3.
For LCMO at x50.3, the role of the t 2g super-exchange seems to be completely screened in terms of the electro-transport and magnetism, or their impact remains far from significant [24][25] . This issue has rarely been questioned in comparison with R 1-x D x CoO 3 where the t 2g exchange becomes important. For example, how do the electrotransport and magnetism respond if the t 2g super-exchange from locally antiferromagnetic (AFM) interaction into locally FM interaction is modulated? In this sense, any approach to enhance the T C via modulating the local t 2g super-exchange would be of interest and highly appreciated not only for practical applications but also for uncovering additional physics of manganites.
In fact, a number of substitution experiments on LCMO by replacing Mn with other transition metal species, e.g. La 12x Ca x Mn 12y Me y O 3 , have been carried out [26][27][28] . Here Me5Fe, Cr, Co, etc and 4d ions in some cases. In most cases the substitutions introduce remarkable quenched disorder and interaction frustrations. For x50.3, these substitutions will lower the FM transition point T C by weakening the double-exchange, and so far no many reports on enhancing the T C in this manganite by means of the Mnsite substitution are available 28 . These experiments may not be unambiguous to uncover the role of the t 2g super-exchange without inducing other complexities. First, the 3d ions may have similar e g levels as those of Mn ion, and thus additional DE effect can't be excluded. In many cases these effects are much more significant than the t 2g -orbital super-exchange [29][30] . So far intention to outshoot the role of the t 2g super-exchange is unsatisfactory. Second, these 3d ions usually have more than one valence state and the substitutions may lead to unexpected variation of Mn valence states and oxygen vacancies 31 . The DE process will be seriously disturbed. Third, these substitutions will induce significant quenched disorder and thus EPS 12,14 .
A handle of these problems is challenging. A promising approach needs to satisfy the following requirements in the low substitution levels. (1) The density of Mn 41 ions as minor chargers should be maintained, so that the density of empty e g orbitals allowing for electron hopping will not be much disturbed. (2) At best the   DE process between Mn ions and substituting ions is prohibited but the t 2g super-exchange can be modulated as much as possible. (3) For most cases the t 2g super-exchanges are AFM. If one is able to introduce an FM super-exchange ingredient at the t 2g orbitals without affecting much the double-exchange, the impact of the t 2g super-exchange definitely deserves for investigation. Along this line, a substitution of Mn by 4d Ru 41 is highly appreciated [26][27] . Before going to details of Ru-substitution in LCMO, we have an outline of www.nature.com/scientificreports Ru-substitution in manganites including the end DMnO 3 systems (x51). In spite of the scattering data, all conclusions state that the Ru-substitution enhances the ferromagnetism. However no detailed discussion on the role of t 2g super-exchange has been addressed regarding this ferromagnetism enhancement. We again take LCMO at x50.3 as an example for illustration of the reasons. Here the Mn-O-Mn bond angle is not so far from 180u [24][25] . This allows a relatively simple scheme of interactions. We consider a neighboring Mn-Ru pair bridged with an O 2ion. The two cases are schematically drawn in Fig. 1(d), (e) where the e g -and t 2g -orbital structures between neighboring Mn 31 -Ru 41 and Mn 41 -Ru 41 pairs are plotted. At the same time, the d-orbital structure for a Ru 41 -Ru 41 pair is shown in Fig. 1(f) for reference, where the t 2g -orbital electron hopping is allowed, as found in SrRuO 3 and CaRuO 3 [32][33][34] . It is noted that Ru ion prefers the Ru 41 valence although other valence states are more or less claimed 35 .
Several aspects of possible physics upon the Ru 41 substitution need to be addressed. First, the e g -orbitals of Ru 41 ion are sufficiently higher than those of Mn ion 36 . This excludes the e g -orbital double-exchange between Mn 31 -Ru 41 pair, as indicated in Fig. 1(d). Second, electron hopping between Ru 41 t 2g -orbitals and Mn e g -orbitals can be questioned too, as indicated in Fig. 1(d) and (e). Instead, the Mn-Ru t 2g super-exchange should be considered. In a good approximation, the t 2g -orbital hybridization and coupling between Mn 31 -Ru 41 and Mn 41 -Ru 41 pairs lead to the FM interaction, which is a critical ingredient of physics we need to consider in this work. In fact, several earlier experiments did reveal the Mn-Ru FM interaction 26,27,29,30,35 , which should be but has not yet been ascribed to this t 2g -orbital FM superexchange. Third, if x is fixed, the Ru-substitution will reduce the d 41 , which certainly damages seriously the Mn 31 -Mn 41 DE sequence and covers up the role of the t 2g super-exchange. A better strategy is to ensure the Mn 31 -Mn 41 DE sequence as disturbed as weak by the Ru substitution. This can be realized in La 0.7-y Ca 0.31y Mn 1-y Ru y O 3 (LCMRO), as long as the d 31 remains sufficiently high, e.g. at y,0.3. It will be shown below that the evaluated Mn-O-Ru bond angle from the structural fitting is ,160u, similar to the Mn-O-Mn bond angle of La 0.7 Ca 0.3 MnO 3 , suggesting that the overlapping between the Mn e g level and Ru t 2g level is weak if any and the scenario shown in Fig. 1(d) and (e) is reasonable. By this scheme, the influence of the Mn 31/41 -Ru 41 t 2g FM super-exchange can be checked by characterizing the magnetism and electro-transport of LCMRO. This is the main motivation of the present work.
Here it should also be mentioned that such a Ru substitution with synchronous Ca substitution in LCMRO will not change the lattice structure much, considering that the ionic sizes of La 31 , Ca 21 , Mn 31 , Mn 41 , and Ru 41 in the 9-coordination frame, are 1.03, 1.00, 0.64, 0.53, and 0.62Å , respectively 37 . Obviously, given roughly constant d 41 , any reduction of the d 31 will damage more or less the electrical conductivity. The quenched disorder and thus the EPS induced by the Ru substitution will be inevitable but negligible, since nearly no thermal hysteresis has been observed for both the magnetization and electrical resistivity measurements. To this stage, we have proposed a scheme for uncovering the effect of the Mn 31/41 -Ru 41 t 2g FM superexchange on the electro-transport and magnetism in LCMRO, and the detailed data are presented below.

Results
Structural characterizations. Both LCMO and CaRuO 3 exhibit orthorhombic structure with space group Pnma 24,32 . Due to similar lattice structures, ionic occupations, and one-to-one corresponding ionic sizes, no serious change in lattice symmetry for the Ru substitution of Mn with synchronous Ca occupation at La site is expected. The measured h22h XRD spectra for a series of samples are presented in Fig. 2(a). The locally amplified reflections around 2h546u ,48u and 68u,70u are presented in Fig. 2(b) and (c), respectively, showing gradually rightward shifting. This is reasonable considering the ionic size mismatch 37 . In addition, the lattice constants (a, b, c) as a function of y respectively, as evaluated by means of the Rietveld refinement processing, are plotted in Fig. 2(d),(f). All the three constants decrease slightly with increasing y, while the unit cell volume V shows a linear dependence on y, satisfying the Vegard's law 38 . This implies that the Ru valence remains to be identical in all the samples, which otherwise would result in an identifiable deviation of the V(y) from the Vegard's law. Here it should be mentioned that the Rietveld refinements confirm the cation ratios for all the samples and one example is given in the Supplementary information.
Further evidence is given by the XPS identification. Fig. 3 presents the XPS data for five samples. In Fig. 3(a) are shown the local nonshifting Ru peaks even when y is as high as 0.5. The XPS spectra covering the 2p 1/2 and 2p 3/2 peaks from an overlap of Mn 31 and Mn 41 are plotted in Fig. 3(b). The overall gradual shifting of the two broad peaks towards the high-energy side with increasing y suggests that the Mn 31 density is gradually lowered. This is strong evidence supporting the gradually lowered Mn 31 density (d 31 ). A rough estimation of the Mn 41 /Mn 31 ratios for several samples in spite of relatively big errors of XPS gives similar results and one example is given in   Surely, the precision of the present XPS data may not be sufficient for excluding the existence of tiny amount of Ru 31 or Ru 51 . However, given the fact that the probed Ru peaks don't move over such a wide y-range from 0.05 to 0.5 but the Mn peaks shift remarkably (see Fig. 3(b)), one is allowed to suggest that the Ru 41 valence state is dominant in these samples even if tiny amount of Ru 31 or Ru 51 is available. In addition, earlier work 39 did report the Ru 31 /Ru 51 valent states in La 0.7 Sr 0.3 Mn 12x Ru x O 3 where the La and Sr contents are constant. However, in our samples, the contents of La, Ca, Mn, and Ru all change synchronously so that the dominant Ru 41 state is favored for the charge neutrality. Furthermore, for the case of high y value (y50.5) where the Ru 31 state was argued to be favored 40 , no peak shift with respect to those samples with lower y can be seen, as shown in Fig. 3.
Magnetic and electro-transport behaviors. In prior to discuss the magnetic and transport data, we check the possible EPS in our samples. It is known that well developed EPS in manganites is usually accompanied with remarkable low-T thermal hysteresis for both the r(T) and M(T) dependences if the measurement is performed in a cooling-warming cycle. We checked all the samples carefully and found no remarked hysteresis and the data for two samples are presented in the Supplementary information. It is seen that the difference between the cooling sequence and warming one is small, suggesting weak EPS if any in these samples.
In the low-y range, an immediate consequence is the gradually damaged electrical conductivity. A reduction of d 31 will certainly dilute the Mn 31 -Mn 41 DE transport networks. To check this effect, we turn to the r(T) and M(T) in response to varying y, plotted in Fig. 4 for y up to 0.50. We mainly discuss the data in the low-y range (y,0.25). For LCMO shown in Fig. 4(a), the r(T) and M(T) dependences reproduce well the earlier data in literature. Upon decreasing T, the MIT at T5T 1 and FM transition at T C occur concurrently with T 1 ,T C ,265K, featured by sharp r(T) peak and M(T) jump from paramagnetic state to FM state 24 . An additional weak bump of the r(T) dependence appears at a lower T,T 2 ,225K, while the overall dependence fits a typical metallic conduction. This bump feature has been well recorded in literature and its origin is believed to be the weak charge-ordering and consequent grain boundaries as weak links for electron transport, resulting in a weak resistivity peak at T 2 41 . We are mainly concerned with the MIT and FM transition at T 1 and T C .
The observed consequences in the low-y range can be described from several aspects. First, the substitution does damage the metallic conduction below the MIT (T 1 ), characterized by a rapid shifting of T 2 towards the low-T side and a slow overall up-rise of r(T) until y,0.10 where the MIT is nearly submerged. As y.0.1, the MIT feature is replaced by an inflexion point (T 1 in Fig. 4(d) and (e)), while the overall behavior is insulator-like. The maximal overall  In parallel to the MIT and FM transition, another consequence of the DE process is the CMR effect around the MIT 15 . The measured MR data at H56T are plotted in Fig. 5. The low-y samples do exhibit the CMR effect with the peak at T,T MR . However, this effect becomes much less pronounced as y.0.15 and nearly disappears at y.0.20, implying that the Mn 31 -Mn 41 double-exchange is suppressed at y.0.20. The observed MR behavior in the high-y range is attributed to the response of the EPS microstructure to increasing H 4 .
As a complimentary, one notices that the rapid decrease of r at y.,0.30 is attributed to the t 2g electron hopping between neighboring Ru 41 -Ru 41 pairs, which becomes remarkable at high y. It is known that CaRuO 3 (SrRuO 3 too) exhibits the metallic conduction due to this strong t 2g electron hopping 32 . Also, the magnetic moment of Ru 41 is much smaller than that of Mn 41 /Mn 3135 , explaining the gradual reduction of M as y.0.30. On the other hand, the substitution will inevitably introduce quenched disorder, and the EPSrelevant phenomena become significant in the high-y range.
Simple percolation model of Mn 31 Fig. 6(a) by the olive color path.
The statistical averaging on these networks indicates a conduction percolation appearing at (d 31 1d 41 ),0.65, i.e. y,0.35, while the probability p for conduction as a function of (d 31 1d 41 ) is plotted in Fig. 6(b). The percolation threshold (y c ,0.35) is roughly consistent with experimentally observed y,0.25 at which the MIT point T 1 disappears, as seen in Fig. 4(e) and Fig. 4(f). In Fig. 6(b) are plotted the simulated results for d 41 50.25 and 0.20, suggesting remarkable dependence of y c on d 41 . Considering the over-simplified DE conduction channels assumed in this toy model lattice as well the negative scattering effects of the less pronounced quenched disorder and EPS on the electron transport, our argument that the electrical conductivity is solely owing to the Mn 31 -Mn 41 double-exchange   is reasonably confirmed, and no substantial contribution from the t 2g -orbital electron hopping is believed.

Discussion
Origin for the unusual ferromagnetism enhancement. The above discussions confirm that the low Ru-substitution with synchronous Ca 21 substitution of La 31 suppresses the DE transport. The major unusual phenomenon is the enhanced FM transition T C . The dependence T C (y) is presented in Fig. 7(a), where the maximal MR temperature T MR , and the MR value at T MR under H56.0T, are plotted together for reference. The r-y data at T5100K and 200K are presented in Fig. 7(b). As revealed earlier, the T C (y) increases linearly in the low-y range (y,0.25) and then falls rapidly at y.0.25, separating the substitution range into two regimes I and II. The r(y) in regime II is dominated by the Ru-Ru interaction, but we only deal with the behaviors in regime I. The r(y) at T5100K and 200K show similar behavior: rapid increase in regime I and then fall in regime II. The T MR (y) dependence is nearly the same as the T C (y) while the MR(y) evidences a rapid fall in regime I, followed by a slow decaying in regime II. This slow decaying is most likely due to the resistivity variations associated with the less pronounced EPS.
To this stage, one can reasonably explain this unusual ferromagnetism enhancement. As shown in Fig. 1(d) and (e), the Mn-Ru FM super-exchange at the t 2g -orbitals is the most probable origin for the ferromagnetism enhancement, while the contribution from the Ru-Ru t 2g FM interaction is negligible in the low-y range. It is noted that the t 2g electrons are usually localized 32 . The Mn-Ru FM interactions at the t 2g -orbitals can strengthen the ferromagnetism but have no benefit to electron hopping in the low-y range, consistent with our observations. To more clearly illustrate this fact, we have to make sure that the interaction between the Mn e g -orbitals and Ru t 2g -orbitals if any should be weak. Since the Ru t 2g -orbital levels are relatively higher than the Mn e g -orbital levels, the Mn e g -orbitals and Ru t 2gorbitals may not have an overlap with each other. A schematic of the arrangement of the Mn d x 2 {y 2 -orbital and Ru d xy -orbital with O 2p-orbitals for a Ru 41 -O-Mn 31 bond is given in Fig. 8(a), noting this bond angle is not very far from 180u. It seems that the effective interaction between the Mn d x 2 {y 2 -orbitals and Ru d xy -orbitals bridged with the O 2p-orbitals is not strong since the Mn-O-Ru bond angle is ,180u in the ideal situation 42 . However, it should be mentioned that the Me-O-Me bond angle here (Me5Mn, Ru) is smaller than 180u, probably allowing a weak electron hopping between the Mn 41 -O 22 -Ru 41 chains. This hopping sequence may contribute somehow to the electrical conduction. In fact, our Rietveld refinements of the XRD data do give a rough estimation of the Mn(2)-O(2)-Ru(1) bond angle to be ,161.71 o for sample y50.05. This angle is already sufficient to avoid the strong interaction between the Mn d x 2 {y 2 -orbitals and Ru d xy -orbitals. It is believed that a much smaller angle is needed for such strong interaction. Therefore, what we need to consider is the super-exchange between the Mn t 2g -orbitals and Ru t 2g -orbitals bridged with the O 2p-orbitals, as shown in Fig. 8(b). One can immediately predict that the super-exchange between them is FM [29][30] and may be sufficiently strong.
To have a semi-quantitative estimation of this Mn-Ru FM super-exchange, we consult recent experimental data in literature.   strong as tens of kelvin 29 . Second, similar Mn-Ru FM effective exchange in manganites can be obtained from the ESR measurements 35 . As mentioned, for y50, the d 31 50.7 is two times higher than the d 41 50.3, the Mn 31 -Mn 41 double-exchange will not be seriously weakened in the low-y range. In the zero-order approximation, the total effective FM exchange may be as big as tens of Kelvin. Therefore, one expects a roughly linear enhancement of T C with y in the low-y range 44 , as revealed in our experiments. Surely, this prediction becomes invalid once y is sufficiently high, e.g. y.0.25 in the present work.
Quenched disorder and electronic phase separation. Finally, we discuss the significance of the possible quenched disorder effects associated with the Ru-substitution with synchronous Casubstitution in LCMRO samples. Anyhow, due to the La 31 -Ca 21 and Mn 31 -Ru 41 ionic size mismatches and charge/orbital fluctuations, the quenched disorder and thus EPS becomes nonnegligible 27 . The anomaly of r(T) at T 2 is somehow attributed to the EPS. We measured the M-H hysteresis loops at T52K and evaluate the saturated magnetization M s and coercivity H c as a function y, as shown in Fig. 9(a) and (b). The rapidly increasing H c (,0.3T at y50.5) can be used to scale the magnitude of the quenched disorder which pins the magnetic domains from switching, noting that CaRuO 3 itself is paramagnetic and SrRuO 3 is ferromagnetic with a coercivity of ,0.1T 32 . At y,0.25, the coercivity is sufficiently small, indicating that the disorder effect if any is weak. On the other hand, it is found that the M s decreases linearly with increasing y, suggesting the rapidly suppressed ferromagnetism in the high-y range.

Methods
The LCMRO polycrystalline samples with x50.3 and y50.0,0.5 were prepared using the convention solid sintering method in air. The highly purified powder of oxides and carbonates was mixed in stoichiometric ratio, ground, and then fired at 1000uC for 24 h in air. The resultant powder was re-ground and pelletized under a pressure of 1000 psi into disks of 2.0 cm in diameter, and then these pellets were sintered at 1300uC for 24 h in air in prior to natural cooling down to room temperature. The chemical composition and spatial homogeneity were checked using the EDS mapping associated with the scanning electron microscopy (SEM, Ultra 55, Zeiss), confirming quite good spatial homogeneity of La, Ca, Mn, and Ru in the mm scale and nano-scale. The evaluated chemical composition is very close to the nominal one within uncertainty of less than 5%. The crystallinity and structure were checked by X-ray diffraction (XRD) using the Cu K a radiation at room temperature with an X-ray power diffractometer (D8 advanced, Bruker). The refinement of the XRD data was performed using the Rietveld method. For checking possible valence state fluctuations upon the Ru substitution, the charge states of Mn and Ru ions were examined by X-ray photoelectron spectroscopy (XPS) using a photon energy of 1253.6 eV (Mg Ka).
We carefully measured the electrical resistivity r and dc magnetization M of the asprepared samples. The M as a function of temperature (T) and magnetic field (H) was measured using the Quantum Design superconducting quantum interference device magnetometer (SQUID) in the zero-field cooled (ZFC) and field-cooling (FC) modes respectively. The cooling field and measuring field were both 100 Oe, sufficiently low so that the magnetic field driven side-effects if any are as weak as possible. In addition, the quasi-static M-H hysteresis loops under high field at different T were measured so that the Ru substitution induced disorder effect can be qualitatively evaluated by characterizing the coercive field (H c ) as a function of the substitution level y. The r(T) and r(H) were measured by a physical properties measurement system (PPMS) from the Cryogenic Co. Ltd and PPMS from the Quantum Design Inc. Since the Mn 31 -Mn 41 double-exchange transport mechanism leads to significant CMR effect occurring around the FM transition point (T C ), we also measured the magnetoresistance MR5[r(0)-r(H)]/r(0). The peak of MR(T) around the T C seems to be a symbol for the significance of the double-exchange transport 4 .